This paper describes an effective method for recovering color signals from multiband images of a high dynamic range (HDR) scene. We note that the color signals in a natural scene have the HDR characteristic of luminance level from very dark shadow area to highly bright sky. The Wiener estimator can be used for estimating spectral-power distributions of the color signals from HDR image data. A previous study presented an improved Wiener estimator for addressing accurate color signal estimation in HDR scenes. However, the previous method required the pixel-by-pixel estimation of parameters contained in the Wiener estimator, which resulted in requiring much computation time. For fast computation, therefore, we propose a lookup-table-based (LUT-based) estimation method for color signals in HDR scenes. In the preliminary stage in advance of color signal estimation, we prepare LUT of the statistical matrix needed in the Wiener estimator, consisting of the covariance matrix of color signals and imaging noises. In the stage of color signal estimation, the estimates are obtained pixel by pixel by the Wiener estimator with the most suitable matrix selected from the LUT. For validating the proposed method, experiments are conducted using actual HDR scenes. Experimental results show the superiority of our method in computation time to the previous methods, with keeping estimation accuracy.
Keita Hirai, Shoji Tominaga, "A LUT-based Method for Recovering Color Signals from High Dynamic Range Images" in Proc. IS&T 20th Color and Imaging Conf., 2012, pp 88 - 93, https://doi.org/10.2352/CIC.2012.20.1.art00016